USE OF PHENOME-WIDE AND GENOME-WIDE APPROACHES TO IDENTIFY PATTERNS OF DISEASE

Abstract

The rise in available longitudinal patient information in electronic health records (EHRs) and their coupling to DNA biobanks has resulted in a dramatic increase in genomic research using EHR data for phenotypic information. Through the use of different phenotyping methods and dissimilar disease processes, we were able to illustrate the strengths and weaknesses of extracting data for research that was curated for clinical medicine. These methods allowed for curation of the largest dataset of individuals with systemic loxoscelism and characterization of this rare disease. This thesis is also the first application of genome-wide risk scores in a phenome-wide approach, demonstrating that genome-wide polygenic risk scores have improved ability to define disease risk and associations. In this body of work, we demonstrated that phenome-wide phenotyping uncertainties can be reduced by grouping of billing codes, large cohort sizes, and the requirement of multiple instances of the billing codes on separate days. Further, we also described an approach in which the results of PheWAS analyses can be validated using multiple cohorts and genomics. These novel methods allowed us to demonstrate the full extent of the role obesity has on postoperative complications and the overall burden of disease driven by obesity in society. Translation of these findings could involve applying genome-wide risk profiling methods to identification of individuals who would benefit from environmental modifications or heightened medical awareness prior to the onset of obesity and its comorbid conditions.